CN113315976A - Three-in-one high information content embedding method for low-resolution video - Google Patents
Three-in-one high information content embedding method for low-resolution video Download PDFInfo
- Publication number
- CN113315976A CN113315976A CN202110588615.0A CN202110588615A CN113315976A CN 113315976 A CN113315976 A CN 113315976A CN 202110588615 A CN202110588615 A CN 202110588615A CN 113315976 A CN113315976 A CN 113315976A
- Authority
- CN
- China
- Prior art keywords
- embedding
- information
- video
- coefficient
- ternary
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 128
- 230000008569 process Effects 0.000 claims abstract description 9
- 229910002056 binary alloy Inorganic materials 0.000 claims abstract description 6
- 230000008878 coupling Effects 0.000 claims description 54
- 238000010168 coupling process Methods 0.000 claims description 54
- 238000005859 coupling reaction Methods 0.000 claims description 54
- 239000011159 matrix material Substances 0.000 claims description 41
- 238000013139 quantization Methods 0.000 claims description 38
- 230000009466 transformation Effects 0.000 claims description 19
- 230000006835 compression Effects 0.000 claims description 15
- 238000007906 compression Methods 0.000 claims description 15
- 238000000605 extraction Methods 0.000 claims description 15
- 230000005540 biological transmission Effects 0.000 claims description 9
- 239000000284 extract Substances 0.000 claims description 8
- 230000003247 decreasing effect Effects 0.000 claims description 4
- 238000012360 testing method Methods 0.000 claims description 4
- 238000004422 calculation algorithm Methods 0.000 abstract description 13
- 238000005516 engineering process Methods 0.000 abstract description 11
- 241000023320 Luma <angiosperm> Species 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- OSWPMRLSEDHDFF-UHFFFAOYSA-N methyl salicylate Chemical compound COC(=O)C1=CC=CC=C1O OSWPMRLSEDHDFF-UHFFFAOYSA-N 0.000 description 4
- 230000003044 adaptive effect Effects 0.000 description 3
- 238000002474 experimental method Methods 0.000 description 3
- 102100029469 WD repeat and HMG-box DNA-binding protein 1 Human genes 0.000 description 2
- 101710097421 WD repeat and HMG-box DNA-binding protein 1 Proteins 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 2
- 230000000903 blocking effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000007667 floating Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- 230000002457 bidirectional effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 210000000887 face Anatomy 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000008520 organization Effects 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/124—Quantisation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/186—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being a colour or a chrominance component
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/46—Embedding additional information in the video signal during the compression process
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
In the prior art, a plurality of H.264 video information hiding algorithms based on a DCT domain exist, but DCT and DST technologies are adopted in an HEVC intra-frame coding process, and all 4 x 4 luminance prediction blocks adopt the DST technology, so that the former H.264 video DCT domain-based video information hiding algorithm is not suitable for HEVC video. The invention provides a ternary information embedding method suitable for low-resolution videos, which replaces the parity binary information embedding method in the prior art, and under the condition that the embedding digits are the same and a unit of 32 bits is used, the information amount is 1.613 times of that of a binary system. The experimental results show that: compared with a binary embedding method, for a QCIF video sequence, the average information embedding amount of the triple-process method is increased by 34.2%, and the PSNR value is basically not reduced.
Description
Technical Field
The invention relates to a video three-in-one high information content embedding method, in particular to a three-in-one high information content embedding method for a low-resolution video, and belongs to the technical field of video high information content embedding.
Background
With the rapid development of the mobile internet, privacy protection and information security are more and more emphasized by people. In recent years, the information hiding technology is rapidly developed, the private information can be transmitted by using popular digital carrier files, the private information can be ensured not to be discovered, and a new generation high efficiency video coding standard (HEVC) comprises the latest coding technology, has excellent compression performance and a special parallel processing design, and is widely applied to various video services.
The high-efficiency video coding HEVC adopts DCT and DST (discrete sine transform), is one of the current advanced video codecs, is a new generation video compression coding standard specified by the International telecommunication organization and the Mobile video experts group, has excellent compression efficiency, is widely applied to the multimedia field (particularly high-definition videos), but has few methods for video information hiding and high information content embedding because the release time of the HEVC video coding standard is not too long.
The coding framework of HEVC coding includes: the method comprises the following steps that firstly, intra-frame prediction is mainly used for removing the spatial correlation of an image, and the current pixel block is predicted to remove spatial redundant information through the information of a reconstructed block after coding, so that the compression efficiency of the image is improved; inter-frame prediction, mainly used for removing the time correlation of the picture, inter-frame prediction obtains the motion information of each block by taking the coded picture as the reference picture of the current frame, removes the time redundancy, improves the compression efficiency, inter-frame prediction in HEVC adopts unidirectional and bidirectional reference pictures for prediction; thirdly, transformation and quantization, namely, the residual data is transformed and quantized to remove frequency domain correlation, the data is compressed in a loss way, the image is transformed from a time domain signal to a frequency domain by transformation and coding, the energy is concentrated to a low-frequency region, the quantization module reduces the dynamic range of image coding, and the transformation and coding and quantization module belong to two mutually independent processes, but in HEVC, the two processes are combined with each other to reduce the calculation complexity; fourthly, deblocking filtering is carried out, blocking effect can occur when a reconstructed image is formed in video coding based on blocks, the blocking effect can be weakened or even eliminated by adopting the deblocking filtering, and the subjective quality and the compression efficiency of the image are improved; fifthly, sample adaptive compensation is carried out, after square block removal filtering is carried out, corresponding offset values are added to pixels by analyzing the statistical characteristics of the pixels subjected to square block removal filtering, the ringing effect is weakened, and the subjective quality and the compression efficiency of the image are improved; and sixthly, entropy coding, namely coding the quantized transform coefficients, the coding control data, the intra-frame prediction data, the motion data and the like into binary streams for storage or transmission.
Compared with the conventional video coding standard, the coding performance of HEVC is greatly improved, which is caused by the use of a new coding tool and a core technology with characteristics. There are many new coding techniques in HEVC, such as adaptive motion vector prediction, AMVP, variable size discrete pre-transform, merging techniques, Merge, mode dependent discrete sine transform, better performing CABAC, and new sample adaptive compensation filters.
The image transformation coding is that pixels of a spatial domain are transformed into a coefficient matrix form after DST transformation, most images contain more flat regions and regions with slow content transformation, the image energy can be transformed into relatively centralized distribution in the transformation domain through proper transformation, spatial redundancy is taken out, effective compression of image information can be obtained by combining other coding technologies such as quantization, z scanning and entropy coding, the HEVC continues to use integers adopted by H.264/AVC and popularizes DCT forms with different sizes, in addition, in order to adapt to distribution conditions of residual errors in different prediction modes, discrete sine transformation DST is introduced into the HEVC, and the integer DST is similar to the integer DCT and also adopted by the HEVC.
HEVC follows the integer DCT technique adopted by h.264/AVC, but its transform matrix is different from that of h.264/AVC, and moreover, HEVC uses 4 different sizes of integer DCTs, 4 × 4, 8 × 8, 16 × 16, and 32 × 32, respectively, and 3 larger sizes of integer DCTs are used in HEVC: 8 × 8, 16 × 16, 32 × 32, the 3 integer DCT derivation methods are substantially the same as 4 × 4, the only difference being the different magnification factor when the matrix elements are integer-quantized.
Compared with h.264/AVC, HEVC has a large improvement in integer DCT, mainly expressed in the following aspects: firstly, HEVC adopts more transforms of larger size, and HEVC mainly faces high-resolution video, which contains more pixels in the same size area, so that the correlation between pixels becomes stronger, and at this time, the compression performance is provided by adopting the transforms of larger size; secondly, integer DCT in HEVC is closer to traditional floating point DCT, HEVC integer DCT matrix retains more precision and is closer to floating point DCT, and better performance can be obtained; and thirdly, the transformation forms of HEVC with different sizes are uniform, and by utilizing the characteristic, a rapid butterfly algorithm with a uniform form can be designed for integer DCT with different sizes.
The HEVC standard uses 4 × 4 integer DST in intra 4 × 4 mode luma component residual coding and integer DCT in intra other modes, all modes between frames, residual coding of all color difference components, which is mainly that intra prediction predicts the current block with surrounding reconstructed block edge pixels so that the intra prediction residual has the following characteristics: the further away from the predicted pixel, the larger the prediction residual amplitude, and the basis functions of DST can adapt well to this feature.
However, there are many disadvantages in the high information content embedding of HEVC video in the prior art, and the difficulties in the prior art and the problems solved by the present invention are mainly focused on the following aspects:
firstly, in the prior art, a plurality of h.264 video information hiding algorithms based on a DCT domain already exist, but DCT and DST technologies are adopted in an HEVC intra-frame coding process, and all 4 × 4 luminance prediction blocks adopt a DST technology, so that the video information hiding algorithm based on the h.264 video DCT domain before is no longer applicable to HEVC video, the prior art lacks an information embedding method applicable to low-resolution video, lacks a better method for replacing parity binary information embedding in the prior art, and the average information embedding amount of the HEVC binary embedding method is too low under the condition that the embedding bits are the same and 32 bits are taken as units, and cannot meet the requirement of high information embedding amount of low-resolution video at all;
secondly, in the prior art, 4 × 4 luminance blocks of the I frame are mostly selected for embedding in the h.264 video information hiding algorithm based on the quantized DST coefficients, because the I frame is a core frame, the quantized coefficients in the B frame and the P frame are fewer than zero coefficients, and the texture of the region is relatively complicated by using 4 × 4 luminance block components during intra-frame prediction, and information embedding is performed in these places, which is difficult for human eyes to find, so the 4 × 4 luminance block is very suitable for embedding video information. However, in the HEVC intra-frame prediction coding, the components of the luminance block include 4 × 4, 8 × 8, 16 × 16, 32 × 32 and 64 × 64, and the HEVC coding standard mainly aims at the compression coding of high-resolution video, and for high-definition super-definition video, in the region with complex texture, a large number of 4 × 4 and 8 × 8 block patterns are adopted during video coding, and in the relatively flat region, a large size block pattern is adopted, but embedding is performed in the flat region, which is easily perceived by human eyes, so that the method is not suitable for information embedding, and the prior art lacks a good solution;
third, the related art transform domain information embedding method generally extracts 1-bit data by the parity of the quantization coefficient, and extracts one-bit data only 0 or 1 due to the high frequency component W33Is 0, if at the time of extraction, W33A value of 0 indicates that the embedded information is 0, W331 indicates that the embedded information is 1, W33A value of-1 indicates that the embedded information is 2, so that embedding one bit of data can express more information than binary, and the coupling coefficient pair group is selected on the basis of texture features
The embedding of the private information can not only thoroughly block the transmission of the embedded error in the block, but also further improve the embedded information quantity.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides a three-step high information content embedding method of low-resolution video, the residual value of a 4X 4 brightness block is subjected to DST transformation, most energy is concentrated in a low-frequency part, and the quantized DST coefficient W of a high-frequency component33Mostly 0, and the correctness of the theory is proved by experiments, and the characteristic is used for proposing a high information content embedding method (W) by using the triple-advance33Values of 0, 1 and-1 represent the embedded information 0, 1 and 2), respectively, in place of the prior art binary parity embedding method, experimental results show that: for QCIF video, the triple-high information content embedding method increases the information content by 34.2% on average, and the PSNR value does not substantially decrease.
In order to achieve the technical effects, the technical scheme adopted by the invention is as follows:
a three-step high information content embedding method for a resolution video is characterized in that a coupling relation exists based on a quantized DST/DST coefficient matrix, namely information is embedded in an embedded coefficient, and a compensation coefficient is correspondingly adjusted to enable some core pixel embedding error values of a reference block to be zero, so that intra-frame distortion drift is relieved, a brightness block with a smaller size is selected for embedding, and 4 × 4 brightness blocks and 8 × 8 brightness blocks which meet conditions are selected for information embedding in a compromise mode;
when the low-resolution video is subjected to predictive coding, a prediction block with the size of 4 x 4 is adopted in a region with relatively complex texture, a prediction block with the size of 8 x 8 is adopted in a relatively flat region, after residual values of a 4 x 4 brightness prediction block are subjected to DST (differential transform and quantization), energy is concentrated in a low-frequency part, high-frequency components are relatively few, and W is relatively low00To W3316 quantization coefficients representing 4 x 4DST, high frequency component W330 is taken as the main component;
the triple-high information content embedding method of the low-resolution video comprises the triple-high information content embedding and the triple-high information content extraction,
the three-step high information content embedding step comprises the following steps:
first, the embedded information is 0, if the quantized coefficient W330, its value is not changed; if the coefficient W is quantized33Is 1, then W is33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if the coefficient W is quantized33Is-1, then W is33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; discarding the embedding if the other value is;
second, embed information as 1 if quantizing coefficient W331, its value is not changed; if the coefficient W is quantized33Is 2, then W33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if the coefficient W is quantized33Is 0, then W is33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; discarding the embedding if the other value is;
third step, the embedded information is 2, if the quantization factor W33If-1, its value is not changed; if the coefficient W is quantized33Is-2, then W33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; if the coefficient W is quantized33Is 0, then W is33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if it is notDiscarding embedding for other values;
the three-step high information quantity extraction step comprises:
step 1, if W33If the bit number is 0, extracting one-bit ternary information 0;
step 2, if W33If the number is 1, extracting one-bit ternary information 1;
step 3, if W33If the value is-1, extracting one-bit ternary information 2;
step 4, if W33And if the value is other, the extraction is not carried out.
The ternary high information content embedding method of the low-resolution video, further, the ternary information embedding zero deduction: an error is generated by embedding private information in a quantization coefficient, and the error is transmitted to a prediction block through a reference block during intra prediction, and in order to effectively reduce the transmission of the error, the error value of a core reference pixel is set to 0, and a quantization DST or DCT coefficient has a coupling relationship, and a coupling coefficient pair is provided (S)1,S2) A pair of quantized DST or DST coefficients in the form of, when paired with S1When the operation of "+ r" is performed, the operation is performed on S2(S2Which may be a vector) to zero the last row or column of matrix B, where S is1For embedding coefficients, S2In order to compensate the coefficient, a residual matrix R of the brightness block at the encoding end is subjected to integer DCT/DST transformation to obtain a quantized DCT/DST coefficient matrix
Wherein QstepIs the quantization step size, whose value is determined by the QP quantization parameter,/represents the dot division operation in a matrix operation, round () is a rounding function.
The method for embedding the ternary high information content of the low-resolution video comprises the following steps of, ternary coupling information coefficient pair: if a certain set of coupled coefficient pairs leads to a quantization systemThe last column of the number matrix is zero, then VS, if a certain set of coupling coefficient pairs makes the last row of the quantization coefficient matrix zero, then HS, resulting in 4X 4 integer DST coupling coefficient pairs and 8X 8 integer DST coupling coefficient pairs, where VS and HS for 4X 4 integer DST are respectively VS and HS
And
VS and HS of 8 by 8 integer DST are respectively
And
the method for embedding the ternary high information content of the low-resolution video comprises the following steps of: in the HEVC intra-frame prediction coding, the luminance block components include 4 × 4, 8 × 8, 16 × 16, 32 × 32 and 64 × 64, and the HEVC coding standard is mainly directed to compression coding of high-resolution video, and for high-definition super-definition video, in areas with complex textures, a large number of 4 × 4 and 8 × 8 block patterns are used in video coding, in relatively flat areas, a large size block pattern is used, but embedding in flat areas is easily perceived by human eyes and is not suitable for information embedding, and embedding in 4 × 4 and 8 × 8 blocks is selected.
The method for embedding the ternary high information content of the low-resolution video comprises the following steps: after the prediction residual value with the size of 4 x 4 is subjected to DST transformation, energy is concentrated on low-frequency components, a high-frequency coefficient is mainly 0, a ternary information embedding method is used for replacing a parity binary information embedding method in the prior art, and under the condition of the same embedding digit, information which can be expressed by the ternary information embedding method is 1.5 times that of the binary information embedding method;
the transform domain information embedding method extracts 1-bit data by the parity of the quantization coefficient, and extracts one-bit data of only 0 or 1 due to the high frequency component W33Is 0, if at the time of extraction, W33A value of 0 indicates that the embedded information is 0, W331 indicates that the embedded information is 1, W33A value of-1 indicates that the embedded information is 2, so that more information can be expressed by embedding one bit of data than binary;
selecting coupling coefficient pair group based on texture characteristics
The embedded information can not only completely block the transmission of embedded errors in the block, but also further improve the embedded information quantity.
The method for embedding the three-step high information content of the low-resolution video comprises the following steps: the intra-frame prediction mode of the adjacent brightness block meets two limiting conditions, the last row or the last column of pixels in the reference block are not used as reference pixels of the prediction block, and if the adjacent brightness block of the current embedded block meets one of the two limiting conditions, information embedding and compensation are carried out in the corresponding coupling factor pair; if the similar vertical texture limiting condition is met, the limiting condition is a first limiting condition; if the horizontal texture limiting condition is met, the limiting condition is a second limiting condition;
the three-step high information content embedding process comprises the following steps: the binary video code stream is partially decoded to obtain a quantization coefficient matrix, if the size is 4 x 4 and the maximum value of the absolute value in the matrix is larger than a critical value R, whether the intra-frame prediction mode restriction condition I or the restriction condition II is met or not is judged, and if the intra-frame prediction mode restriction condition I or the restriction condition II is met, 1-bit ternary information is embedded in the corresponding coupling coefficient pair.
The three-step high information content embedding method of the low-resolution video comprises the following steps of: the binary video code stream is partially decoded to obtain a quantization coefficient matrix, if the size is 4 x 4 and the maximum value of the absolute value in the matrix is larger than a critical value R, whether the intra-frame prediction mode restriction condition I or the restriction condition II is met or not is judged, and if the intra-frame prediction mode restriction condition I or the restriction condition II is met, 1-bit ternary information is extracted from the corresponding coupling coefficient pair.
The method for embedding the high information content of the triple system of the low-resolution video is further realized on HEVC video standard reference software HM, video sequences YUV are tested, the video sequences are CREW, CITY, HARBOUR and SOCCER, the resolution of each video sequence is 176 multiplied by 144-QCIF, 352 multiplied by 288-CIF and 704 multiplied by 576-4CIF respectively, wherein the frame rate of the video sequences with the resolutions of 176 multiplied by 144 and 352 multiplied by 288 is 15 frames per second, and the resolution is 704 multiplied by 576 video sequences is 30 frames per second; the coding interval is 4, each video sequence has 25I frames, and the quantization step length of each I frame is 28;
PSNR1 represents a peak signal-to-noise ratio calculated by comparing a video after embedding private information with an original YUV video before encoding, PSNR2 represents a peak signal-to-noise ratio calculated by comparing a video after embedding private information with a video after decoding an HEVC video code stream, PSNR1 and PSNR2 are average values of the peak signal-to-noise ratios of all frames, embedding capacity represents the number of bits of each I frame for average embedding, R is a custom critical value, when the critical value R is 0, the PSNR1 values before and after embedding are reduced by 0.1213 to 0.4621dB, and the average I frame embedding capacity is 131 to 1221 bits.
The three-step high information content embedding method of low resolution video is characterized by that it compares the three-step high information content embedding method based on 4X 4 brightness block with binary information embedding method of existent technology, and its embedding capacity is less than that of binary embedding method, and the ternary only selects 4X 4 brightness block which is in accordance with condition to embed, and does not select 8X 8 brightness block to embed, and the ternary embedding one-bit information to be expressed is more than binary, and takes 32bit binary information as example, its expressible number is at most 232-1, and its expression by ternary only needs 19, and 3 can be expressed by ternary21-1＞232-1, in the same number of embedding bits, the information amount that can be expressed by the ternary system is 32/19 times binary 32/19, 231 bits in city (qcif) binary embedding amount and 211 bits in ternary system, so that the information amount expressed by the ternary system is 211 × 1.5/231 times 1.37, the information amount of the ternary system is increased by 37% compared with the information amount of the binary system, and the PSNR1 is reduced very little and is within 0.1 dB.
The invention relates to a three-step high information content embedding method for low-resolution video, in particular to a three-step high information content embedding method for low-resolution video, which compares the information content increasing proportion of a three-step embedding method (QCIF, CIF and 4CIF) relative to a binary embedding method under different resolutions, the information content increasing proportion of the same video sequence is reduced along with the increase of the resolution, even the information content is reduced, 8X 8 brightness blocks are more and more along with the increase of the resolution, and only 4X 4 blocks are selected for embedding by the three-step embedding method.
Compared with the prior art, the invention has the following contributions and innovation points:
first, there are many h.264 video information concealment algorithms based on the DCT domain in the prior art, but the DCT and DST techniques are adopted in the HEVC intra-frame coding process, and all 4 × 4 luminance prediction blocks adopt the DST technique, which makes the previous h.264 video DCT domain based video information concealment algorithm no longer applicable to HEVC video. The invention provides a ternary information embedding method suitable for low-resolution videos, which replaces the parity binary information embedding method in the prior art, and under the condition that the embedding digits are the same and a unit of 32 bits is used, the information amount is 1.613 times of that of a binary system. The experimental results show that: compared with a binary embedding method, for a QCIF video sequence, the average information embedding amount of a triple-processing method is increased by 34.2%, and the PSNR value is basically not reduced, so that the triple-processing high information embedding method is applied to a low-resolution video, and has high commercial value and wide market application prospect;
secondly, when the low-resolution video is subjected to predictive coding, a prediction block with the size of 4 × 4 is adopted in a region with relatively complex texture, and a prediction block with the size of 8 × 8 is adopted in a relatively gentle region, so that information embedding in an 8 × 8 block easily causes subjective visual attention and affects the safety of video embedded information, the energy of a prediction residual value with the size of 4 × 4 is concentrated in a low-frequency component after the DST transformation, the high-frequency coefficient is mostly 0, a ternary information embedding method is used for replacing a parity binary information embedding method in the prior art, and under the condition of the same embedding digit, the information which can be expressed by the ternary information embedding method is about 1.5 times that of the binary information embedding method, so that the transmission of an embedded error in the block can be blocked, and the embedded information amount is further improved;
thirdly, comparing the three-step high information embedding method based on 4 x 4 brightness block with the binary information embedding method of the prior art, the three-step high information embedding method has less embedding capacity than the binary embedding method because the ternary only selects the qualified 4 x 4 brightness block for embedding, but does not select the 8 x 8 brightness block for embedding, and the ternary embedding one bit of information to be expressed is more than the binary, taking 32bit binary information as an example, the number which can be expressed is at most 232-1, and the expression by the ternary only needs 19 for expression, 321-1＞2321, in the same number of embedded bits, the amount of information that can be expressed by the ternary system is 32/19 times binary 32/19, 231 bits in city (qcif) binary embedded amount and 211 bits in ternary system, so that the amount of information expressed by the ternary system is 211 × 1.5/231 times 1.37, the amount of embedded information in ternary system is increased by 37%, and PSNR1 is reduced little and within 0.1 dB;
fourthly, comparing the information quantity increasing proportion of the ternary embedding method relative to the binary embedding method under different resolutions (QCIF, CIF and 4CIF), it can be seen that the information quantity increasing proportion of the same video sequence is reduced along with the increase of the resolution, and even the information quantity is reduced (video sequence creew (4CIF)), because as the resolution is increased, 8 x 8 luminance blocks are more and more, and the ternary embedding method only selects 4 x 4 blocks for embedding.
Drawings
Fig. 1 is a diagram of a quantized DST coefficient matrix according to the present invention.
Fig. 2 is an embedding flow chart of the three-step high information content embedding method of the present invention.
FIG. 3 is a flow chart of the extraction of the three-step high information content embedding method of the present invention.
Fig. 4 is a schematic diagram of objective evaluation indexes of the ternary video information concealment algorithm according to the present invention.
FIG. 5 is a graph comparing the indicators of the ternary high information content embedding method and the binary embedding method of the present invention.
Detailed Description
The technical solution of the triple-high information content embedding method for low resolution video provided by the present invention is further described below with reference to the accompanying drawings, so that those skilled in the art can better understand the present invention and can implement the same.
In the prior art, a plurality of h.264 video information hiding algorithms based on the DCT domain already exist, but DCT and DST technologies are adopted in the HEVC intra-frame coding process, and all 4 × 4 luminance prediction blocks adopt the DST technology, so that the conventional h.264 video DCT domain based video information hiding algorithm is not suitable for HEVC video. The invention provides a ternary information embedding method suitable for low-resolution videos, which replaces the parity binary information embedding method in the prior art, and under the condition that the embedding digits are the same and a unit of 32 bits is used, the information amount is 1.613 times of that of a binary system. The experimental results show that: compared with the binary embedding method, the average information embedding amount of the ternary embedding method is increased by 34.2% for the QCIF video sequence, and the PSNR value is not reduced basically.
One, three system information embedding zero-return inference
Embedding of private information in the quantized coefficients causes errors, and the errors are transferred to the predicted block by the reference block during intra prediction and accumulated. To solve this problem, the error propagation is effectively reduced, the error value of the core reference pixel is set to 0, the quantized DST or DCT coefficient has a coupling relationship, and the coupling coefficient pair is provided with (S)1,S2) A pair of quantized DST or DST coefficients in the form of, when paired with S1When the operation of "+ r" is performed, the operation is performed on S2(S2May be a vector) may be zeroed out for the last row or column of matrix B, where S is zero, for an "-r" or "+ r" operation1For embedding coefficients, S2Is a compensation factor. After the residual error matrix R of the brightness block at the encoding end is subjected to integer DCT/DST transformation, a quantized DCT/DST coefficient matrix is obtained
Wherein QstepIs the quantization step size, whose value is determined by the QP quantization parameter,/represents the dot division operation in a matrix operation, round () is a rounding function.
Binary and ternary coupling information coefficient pair
If a certain group of coupling coefficient pairs makes the last column of the quantized coefficient matrix zero, then VS is defined as VS, if a certain group of coupling coefficient pairs makes the last column of the quantized coefficient matrix zero, then HS, resulting in 4X 4 integer DST coupling coefficient pairs and 8X 8 integer DST coupling coefficient pairs, where VS and HS of 4X 4 integer DST are VS and HS respectively
And
VS and HS of 8 by 8 integer DST are respectively
And
selection of ternary or ternary embedded blocks
In the prior art, 4 × 4 luminance blocks of an I frame are mostly selected to be embedded in a video information hiding algorithm based on quantized DST coefficients, because the I frame is a core frame, quantized coefficients in a B frame and a P frame are few, and a 4 × 4 luminance block component is adopted to represent that the texture of the region is relatively complex during intra-frame prediction, information embedding is carried out in the positions, and the human eyes can hardly find the positions, so that the 4 × 4 luminance block is very suitable for embedding video information. However, in HEVC intra-frame prediction coding, there are 4 × 4, 8 × 8, 16 × 16, 32 × 32, and 64 × 64 components of a luminance block, and the HEVC coding standard is mainly directed to compression coding of high-resolution video, and for high-definition super-definition video, in an area with a complicated texture, a large number of 4 × 4 and 8 × 8 block patterns are used in video coding, and in a relatively flat area, a relatively large size block pattern is used, but embedding in a flat area is easily perceived by the naked eye, and therefore, it is not suitable for information embedding. Based on the above considerations, 4 × 4 and 8 × 8 blocks were selected for embedding.
The continuous emergence of video application puts forward higher requirements on video coding, and further pushes the video coding standard to advance to higher compression efficiency, and HEVC has quite excellent compression performance, so that it will play a great role in high-definition super-definition, 2K and 4K video applications. The invention starts from HEVC intra-frame prediction coding, analyzes the reason of embedding error caused by embedding private information in a quantization coefficient matrix, and because partial pixels (the last row and the last column) of a coded brightness block can be used as reference pixels of an uncoded brightness block, the error can be transmitted along with the transmission of the error, intra-frame distortion drift is generated, in order to effectively control the intra-frame distortion drift, coupling relation exists based on the quantization DST/DST coefficient matrix, namely information is embedded in an embedding coefficient, and the compensation coefficient is correspondingly adjusted, so that the embedding error value of some core pixels of a reference block is zero, the intra-frame distortion drift is relieved, according to the characteristics of human eyes, information embedding is carried out in a region with more complex image details, so that the human eyes can hardly distinguish, a brightness block with smaller size is selected for embedding, but because the mode of the large-size block in a high-definition video is larger, the trade-off therefore considers the selection of eligible 4 x 4 luma blocks and 8 x 8 luma blocks for privacy information embedding.
Four, three system information embedding method
When the low-resolution video is subjected to predictive coding, a prediction block with the size of 4 x 4 is adopted in a region with relatively complex texture, a prediction block with the size of 8 x 8 is adopted in a relatively gentle region, so that information embedding in the 8 x 8 block easily attracts subjective visual attention and influences the safety of video embedding information, after the prediction residual value with the size of 4 x 4 is subjected to DST transformation, energy is concentrated in low-frequency components, most of high-frequency coefficients are 0, a ternary information embedding method is used for replacing a parity binary information embedding method in the prior art, and under the condition of the same embedding digit number, the information which can be expressed by the ternary information embedding method is about 1.5 times that of the binary information embedding method.
Ternary (one) information embedding analysis
After the residual values of the 4 × 4 luma prediction block are quantized by the DST transform, the energy is concentrated in the low frequency part, and the high frequency components are relatively small, as shown in fig. 1, W00To W3316 quantization coefficients representing 4 x 4DST, high frequency component W33Most of the time is 0.
The invention tests W in the quantized DST matrix33The number of absolute values greater than 1 accounts for the percentage of the total number, and the experimental environment is as follows: the quantization step size of the I frames is respectively 28, the tested video sequences are CREW, CITY, HARBOUR and SOCCER, the resolution of each video sequence is respectively 176 x 144-QCIF, 352 x 288-CIF and 704 x 576-4CIF, the coding interval is 4, each video sequence has 25I frames, and the experimental result is that: video sequence city (QCIF), W33The number of absolute values greater than 1 is 3, and the total number is 20856; video sequence City (CIF), W33The number of absolute values greater than 1 is 5, and the total number is 67608; video sequence city (4CIF), W33The number of absolute values greater than 1 is 3, and the total number is 132372; video sequence (QCIF), W33The number of absolute values greater than 1 is 2, and the total number is 16884; video sequence Crop (CIF), W33The number of absolute values greater than 1 is 2, and the total number is 50164; video sequence crop (4CIF), W33The number of absolute values greater than 1 is 2, and the total number is 87660; video sequence harbor (QCIF), W33The number of absolute values greater than 1 is 1, and the total number is 22892; video sequence harbor (CIF), W33The number of absolute values greater than 1 is 2, and the total number is 68688; video sequence harbor (4CIF), W33The number of absolute values greater than 1 is 0, and the total number is 80572; video sequence soccer (QCIF), W33The number of absolute values greater than 1 is 6, and the total number is 12612; video sequence soccer (CIF), W33The number of absolute values greater than 1 is 9, and the total number is 28292; video sequence soccer (4CIF), W33The number of absolute values greater than 1 is 0, and the total number is 60076; w33The number of absolute values greater than 1 is very small, almostCan be ignored.
The prior art transform domain information embedding method usually extracts 1-bit data by the parity of the quantization coefficient, and extracts one-bit data only 0 or 1 due to the high frequency component W33Is 0, if at the time of extraction, W33A value of 0 indicates that the embedded information is 0, W331 indicates that the embedded information is 1, W33A value of-1 indicates that the embedded information is 2, so that more information can be expressed by embedding one bit of data than binary.
Selecting coupling coefficient pair group based on texture characteristics
The embedding of the private information can not only thoroughly block the transmission of the embedded error in the block, but also further improve the embedded information quantity.
(II) three-in high information content embedding and extracting process
1. Triple-feed high information content embedding
The intra-frame prediction mode of the adjacent brightness block meets two limiting conditions, the last row or the last column of pixels in the reference block are not used as reference pixels of the prediction block, and if the adjacent brightness block of the current embedded block meets one of the two limiting conditions, information embedding and compensation are carried out in the corresponding coupling factor pair; if the similar vertical texture limiting condition is met, the limiting condition is a first limiting condition; and if the horizontal texture constraint condition is met, the constraint condition is a second constraint condition.
The flow chart of embedding the ternary high information content is shown in fig. 2, a quantized coefficient matrix is obtained after a binary video code stream is partially decoded, if the size is 4 x 4 and the maximum absolute value in the matrix is greater than a critical value R, whether the intra-frame prediction mode restriction condition one or the restriction condition two is met is judged, and if the size is met, 1-bit ternary information is embedded in a corresponding coupling coefficient pair.
The three-step high information content embedding step comprises the following steps:
first, the embedded information is 0, if the quantized coefficient W330, its value is not changed; if the coefficient W is quantized33Is 1, then W is33Subtracting 1 and adjusting the coupling accordinglyThe values of the other coefficients in the pair of coefficients; if the coefficient W is quantized33Is-1, then W is33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; discarding the embedding if the other value is;
second, embed information as 1 if quantizing coefficient W331, its value is not changed; if the coefficient W is quantized33Is 2, then W33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if the coefficient W is quantized33Is 0, then W is33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; discarding the embedding if the other value is;
third step, the embedded information is 2, if the quantization factor W33If-1, its value is not changed; if the coefficient W is quantized33Is-2, then W33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; if the coefficient W is quantized33Is 0, then W is33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if it is the other value, the embedding is discarded.
2. Three-step high information quantity extraction
The flow chart of extracting the ternary high information content is shown in fig. 3, a quantized coefficient matrix is obtained after a binary video code stream is partially decoded, if the size is 4 x 4 and the maximum absolute value in the matrix is greater than a critical value R, whether the intra-frame prediction mode restriction condition one or the restriction condition two is met is judged, and if the size is 4 x 4 and the maximum absolute value in the matrix is greater than the critical value R, 1-bit ternary information is extracted from a corresponding coupling coefficient pair.
The three-step high information quantity extraction step comprises:
step 1, if W33If the bit number is 0, extracting one-bit ternary information 0;
step 2, if W33If the number is 1, extracting one-bit ternary information 1;
step 3, if W33If the value is-1, extracting one-bit ternary information 2;
step 4, if W33And if the value is other, the extraction is not carried out.
(III) analysis of the results of the experiment
The method is realized on HEVC video standard reference software HM, and the experimental environment is as follows: testing video sequences YUV, wherein the testing video sequences are CREW, CITY, HARBOUR and SOCCER, the resolution of each video sequence is 176 multiplied by 144-QCIF, 352 multiplied by 288-CIF and 704 multiplied by 576-4CIF respectively, the frame rate of the video sequences with the resolution of 176 multiplied by 144 and 352 multiplied by 288 is 15 frames per second, and the resolution of the video sequence with the resolution of 704 multiplied by 576 is 30 frames per second; the coding interval is 4, each video sequence has 25I frames, and the quantization step size of each I frame is 28.
PSNR1 represents the peak snr calculated by comparing the video after embedding private information with the original YUV video before encoding, PSNR2 represents the peak snr calculated by comparing the video after embedding private information with the video after decoding the HEVC video stream, PSNR1 and PSNR2 are both the average of the peak snrs of all frames, the embedding capacity represents the number of bits of each I frame to be embedded on average, R is a custom threshold, the objective evaluation index of the ternary-based video information concealment algorithm is shown in fig. 4, when the threshold R is 0, the PSNR1 values before and after embedding are reduced by 0.1213 to 0.4621dB, and the average I frame embedding capacity is 131 to 1221 bits.
The three-step high information content embedding method based on 4 x 4 luminance blocks is compared with the binary information embedding method of the prior art, and the comparison result is shown in fig. 5. The embedding capacity of the ternary high information quantity embedding method is less than that of the binary embedding method, because the ternary only selects the 4X 4 brightness blocks which meet the conditions for embedding, but does not select the 8X 8 brightness blocks for embedding, and the ternary embedding one bit of information to be expressed is more than that of the binary, taking 32bit binary information as an example, the number which can be expressed is at most 232-1, and the expression by the ternary can be expressed by only 19, 321-1＞232-1, in the same number of embedding bits, the information amount that can be expressed by the ternary system is 32/19 times binary 32/19, 231 bits in city (qcif) binary embedding amount and 211 bits in ternary system, so that the information amount expressed by the ternary system is 211 × 1.5/231 times 1.37, the information amount of the ternary system is increased by 37% compared with the information amount of the binary system, and the PSNR1 is reduced very little and is within 0.1 dB.
Comparing the information amount increasing ratios of the ternary embedding method to the binary embedding method under different resolutions (QCIF, CIF and 4CIF), it can be seen that the information amount increasing ratio is decreasing with the increase of the resolution of the same video sequence, and even the information amount is decreasing (the video sequence creew (4CIF)) because as the resolution is increased, 8 × 8 luminance blocks are more and more, and the ternary embedding method only selects 4 × 4 blocks for embedding. In summary, the triple-rate high-throughput embedding method of the present invention is particularly suitable for embedding information in low-resolution video, and the average increased throughput rate is 34.2% for QCIF video sequences.
After the 4 x 4 luminance block residual value is subjected to the DST transformation, most of energy is concentrated in a low-frequency part, and quantized DST coefficients W of a high-frequency component33Mostly 0, and the correctness of the theory is proved by experiments, and the characteristic is used for proposing a high information content embedding method (W) by using the triple-advance33Values of 0, 1 and-1 represent the embedded information 0, 1 and 2), respectively, in place of the prior art binary parity embedding method, experimental results show that: for QCIF video, the triple-high information content embedding method increases the information content by 34.2% on average, and the PSNR value does not substantially decrease.
Claims (10)
1. The three-step high information content embedding method of the low-resolution video is characterized in that coupling relation exists based on a quantized DST/DST coefficient matrix, namely information is embedded in an embedded coefficient, and a compensation coefficient is correspondingly adjusted to enable embedding error values of some core pixels of a reference block to be zero, so that intra-frame distortion drift is relieved, a brightness block with a smaller size is selected for embedding, and 4 × 4 brightness blocks and 8 brightness blocks which meet conditions are selected for information embedding in a compromise mode;
when the low-resolution video is subjected to predictive coding, a prediction block with the size of 4 x 4 is adopted in a region with relatively complex texture, a prediction block with the size of 8 x 8 is adopted in a relatively flat region, after residual values of a 4 x 4 brightness prediction block are subjected to DST (differential transform and quantization), energy is concentrated in a low-frequency part, high-frequency components are relatively few, and W is relatively low00To W3316 quantization coefficients representing 4 x 4DST, high frequency component W330 is taken as the main component;
the triple-high information content embedding method of the low-resolution video comprises the triple-high information content embedding and the triple-high information content extraction,
the three-step high information content embedding step comprises the following steps:
first, the embedded information is 0, if the quantized coefficient W330, its value is not changed; if the coefficient W is quantized33Is 1, then W is33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if the coefficient W is quantized33Is-1, then W is33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; discarding the embedding if the other value is;
second, embed information as 1 if quantizing coefficient W331, its value is not changed; if the coefficient W is quantized33Is 2, then W33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; if the coefficient W is quantized33Is 0, then W is33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; discarding the embedding if the other value is;
third step, the embedded information is 2, if the quantization factor W33If-1, its value is not changed; if the coefficient W is quantized33Is-2, then W33Adding 1 and correspondingly adjusting the values of other coefficients in the coupling coefficient pair; if the coefficient W is quantized33Is 0, then W is33Subtracting 1 and adjusting the values of the other coefficients in the coupling coefficient pair accordingly; discarding the embedding if the other value is;
the three-step high information quantity extraction step comprises:
step 1, if W33If the bit number is 0, extracting one-bit ternary information 0;
step 2, if W33If the number is 1, extracting one-bit ternary information 1;
step 3, if W33If the value is-1, extracting one-bit ternary information 2;
step 4, if W33And if the value is other, the extraction is not carried out.
2. Ternary for low resolution video according to claim 1The high information content embedding method is characterized in that ternary information embedding zeroing inference: an error is generated by embedding private information in a quantization coefficient, and the error is transmitted to a prediction block through a reference block during intra prediction, and in order to effectively reduce the transmission of the error, the error value of a core reference pixel is set to 0, and a quantization DST or DCT coefficient has a coupling relationship, and a coupling coefficient pair is provided (S)1,S2) A pair of quantized DST or DST coefficients in the form of, when paired with S1When the operation of "+ r" is performed, the operation is performed on S2(S2Which may be a vector) to zero the last row or column of matrix B, where S is1For embedding coefficients, S2In order to compensate the coefficient, a residual matrix R of the brightness block at the encoding end is subjected to integer DCT/DST transformation to obtain a quantized DCT/DST coefficient matrix
Wherein QstepIs the quantization step size, whose value is determined by the QP quantization parameter,/represents the dot division operation in a matrix operation, round () is a rounding function.
3. The method of claim 1, wherein the information coefficient pairs of ternary coupling are: if a certain group of coupling coefficient pairs makes the last column of the quantized coefficient matrix zero, then VS is defined as VS, if a certain group of coupling coefficient pairs makes the last column of the quantized coefficient matrix zero, then HS, resulting in 4X 4 integer DST coupling coefficient pairs and 8X 8 integer DST coupling coefficient pairs, where VS and HS of 4X 4 integer DST are VS and HS respectively
And
VS and HS of 8 by 8 integer DST are respectively
And
(i＝0,1,2...7)。
4. the method of claim 1, wherein the ternary embedded block is selected from the group consisting of: in the HEVC intra-frame prediction coding, the luminance block components include 4 × 4, 8 × 8, 16 × 16, 32 × 32 and 64 × 64, and the HEVC coding standard is mainly directed to compression coding of high-resolution video, and for high-definition super-definition video, in areas with complex textures, a large number of 4 × 4 and 8 × 8 block patterns are used in video coding, in relatively flat areas, a large size block pattern is used, but embedding in flat areas is easily perceived by human eyes and is not suitable for information embedding, and embedding in 4 × 4 and 8 × 8 blocks is selected.
5. The method of claim 1, wherein the ternary information embedding method comprises: after the prediction residual value with the size of 4 x 4 is subjected to DST transformation, energy is concentrated on low-frequency components, a high-frequency coefficient is mainly 0, a ternary information embedding method is used for replacing a parity binary information embedding method in the prior art, and under the condition of the same embedding digit, information which can be expressed by the ternary information embedding method is 1.5 times that of the binary information embedding method;
the transform domain information embedding method extracts 1-bit data by the parity of the quantization coefficient, and extracts one-bit data of only 0 or 1 due to the high frequency component W33Is 0, if at the time of extraction, W33A value of 0 indicates that the embedded information is 0, W331 indicates that the embedded information is 1, W33A value of-1 indicates that the embedded information is 2, so that more information can be expressed by embedding one bit of data than binary;
based on texture featuresOn the basis of the above-mentioned data, selecting coupling coefficient pair group
The embedded information can not only completely block the transmission of embedded errors in the block, but also further improve the embedded information quantity.
6. The method of claim 1, wherein the high-resolution video is embedded in a ternary manner by: the intra-frame prediction mode of the adjacent brightness block meets two limiting conditions, the last row or the last column of pixels in the reference block are not used as reference pixels of the prediction block, and if the adjacent brightness block of the current embedded block meets one of the two limiting conditions, information embedding and compensation are carried out in the corresponding coupling factor pair; if the similar vertical texture limiting condition is met, the limiting condition is a first limiting condition; if the horizontal texture limiting condition is met, the limiting condition is a second limiting condition;
the three-step high information content embedding process comprises the following steps: the binary video code stream is partially decoded to obtain a quantization coefficient matrix, if the size is 4 x 4 and the maximum value of the absolute value in the matrix is larger than a critical value R, whether the intra-frame prediction mode restriction condition I or the restriction condition II is met or not is judged, and if the intra-frame prediction mode restriction condition I or the restriction condition II is met, 1-bit ternary information is embedded in the corresponding coupling coefficient pair.
7. The method of claim 1, wherein the tri-system high information content extraction: the binary video code stream is partially decoded to obtain a quantization coefficient matrix, if the size is 4 x 4 and the maximum value of the absolute value in the matrix is larger than a critical value R, whether the intra-frame prediction mode restriction condition I or the restriction condition II is met or not is judged, and if the intra-frame prediction mode restriction condition I or the restriction condition II is met, 1-bit ternary information is extracted from the corresponding coupling coefficient pair.
8. The method of claim 1, wherein the method is implemented in HEVC video standard reference software HM, the video sequences YUV, the test video sequences CREW, CITY, HARBOUR, SOCCER, the resolution of each video sequence being 176 × 144-QCIF, 352 × 288-CIF, and 704 × 576-4CIF, respectively, wherein the video sequence frame rates with the resolutions 176 × 144, 352 × 288 are 15 frames per second, and the video sequence frame rate with the resolution 704 × 576 is 30 frames per second; the coding interval is 4, each video sequence has 25I frames, and the quantization step length of each I frame is 28;
PSNR1 represents a peak signal-to-noise ratio calculated by comparing a video after embedding private information with an original YUV video before encoding, PSNR2 represents a peak signal-to-noise ratio calculated by comparing a video after embedding private information with a video after decoding an HEVC video code stream, PSNR1 and PSNR2 are average values of the peak signal-to-noise ratios of all frames, embedding capacity represents the number of bits of each I frame for average embedding, R is a custom critical value, when the critical value R is 0, the PSNR1 values before and after embedding are reduced by 0.1213 to 0.4621dB, and the average I frame embedding capacity is 131 to 1221 bits.
9. The method of claim 8, wherein the tri-row high information content embedding method based on 4 x 4 luminance blocks is compared with the prior art binary information embedding method, the tri-row high information content embedding method has less embedding capacity than the binary embedding method, the tri-row method selects only the qualified 4 x 4 luminance blocks for embedding, but not the 8 x 8 luminance blocks for embedding, and the tri-row embedding one bit of information to be expressed is more than the binary, taking 32bit binary information as an example, the number capable of expressing is at most 232-1, and the expression by the tri-row method only needs 19, and 3 can express the information by the tri-row method, 321-1＞232-1, in the same number of embedding bits, the information amount that can be expressed by the ternary system is 32/19 times binary 32/19, 231 bits in city (qcif) binary embedding amount and 211 bits in ternary system, so that the information amount expressed by the ternary system is 211 × 1.5/231 times 1.37, the information amount of the ternary system is increased by 37% compared with the information amount of the binary system, and the PSNR1 is reduced very little and is within 0.1 dB.
10. The method according to claim 8, wherein the rate of increase of the information content of the ternary embedding method relative to the binary embedding method is compared at different resolutions (QCIF, CIF and 4CIF), and the same video sequence is decreased with the increase of the resolution and even decreased with the increase of the information content, and as the resolution is increased, 8 × 8 luminance blocks are more and more, and the ternary embedding method selects only 4 × 4 blocks for embedding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110588615.0A CN113315976A (en) | 2021-05-28 | 2021-05-28 | Three-in-one high information content embedding method for low-resolution video |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110588615.0A CN113315976A (en) | 2021-05-28 | 2021-05-28 | Three-in-one high information content embedding method for low-resolution video |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113315976A true CN113315976A (en) | 2021-08-27 |
Family
ID=77375716
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110588615.0A Pending CN113315976A (en) | 2021-05-28 | 2021-05-28 | Three-in-one high information content embedding method for low-resolution video |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113315976A (en) |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101640802A (en) * | 2009-08-28 | 2010-02-03 | 北京工业大学 | Video inter-frame compression coding method based on macroblock features and statistical properties |
| CN104050622A (en) * | 2014-06-02 | 2014-09-17 | 鲁东大学 | Color image blind-watermarking method based on ternary coding |
| US20140278446A1 (en) * | 2013-03-18 | 2014-09-18 | Fujitsu Limited | Device and method for data embedding and device and method for data extraction |
| CN107172434A (en) * | 2017-06-27 | 2017-09-15 | 郑州师范学院 | A kind of method that built-in steganography is carried out to H.265 video |
| CN109525845A (en) * | 2018-12-04 | 2019-03-26 | 郑州师范学院 | A kind of built-in Information hiding verification method suitable for video |
| CN109803143A (en) * | 2018-10-16 | 2019-05-24 | 曙光云计算集团有限公司 | A kind of pair of video carries out the hiding method of frame information and device |
-
2021
- 2021-05-28 CN CN202110588615.0A patent/CN113315976A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101640802A (en) * | 2009-08-28 | 2010-02-03 | 北京工业大学 | Video inter-frame compression coding method based on macroblock features and statistical properties |
| US20140278446A1 (en) * | 2013-03-18 | 2014-09-18 | Fujitsu Limited | Device and method for data embedding and device and method for data extraction |
| CN104050622A (en) * | 2014-06-02 | 2014-09-17 | 鲁东大学 | Color image blind-watermarking method based on ternary coding |
| CN107172434A (en) * | 2017-06-27 | 2017-09-15 | 郑州师范学院 | A kind of method that built-in steganography is carried out to H.265 video |
| CN109803143A (en) * | 2018-10-16 | 2019-05-24 | 曙光云计算集团有限公司 | A kind of pair of video carries out the hiding method of frame information and device |
| CN109525845A (en) * | 2018-12-04 | 2019-03-26 | 郑州师范学院 | A kind of built-in Information hiding verification method suitable for video |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR100772883B1 (en) | Deblocking filtering method considering intra BL mode, and video encoder/decoder based on multi-layer using the method | |
| CN111819854B (en) | Method and apparatus for coordinating multi-sign bit concealment and residual sign prediction | |
| WO2006041305A1 (en) | Deblocking filter | |
| US7822125B2 (en) | Method for chroma deblocking | |
| KR20130045154A (en) | Image decoding apparatus | |
| JP2003521178A (en) | Quality-based image compression | |
| US20110194615A1 (en) | Video sequence encoding system and algorithms | |
| WO2008118562A1 (en) | Simplified deblock filtering for reduced memory access and computational complexity | |
| WO2010078759A1 (en) | Method for image temporal and spatial resolution processing based on code rate control | |
| KR20060080107A (en) | Deblocking filtering method considering intra bl mode, and video encoder/decoder based on multi-layer using the method | |
| KR20130045153A (en) | Image decoding apparatus | |
| KR20130011878A (en) | Image encoding/decoding method and apparatus using deblocking filtering | |
| KR20190091422A (en) | Method and apparatus for generating reconstruction block | |
| CA2533885A1 (en) | Method and apparatus for selection of scanning mode in dual pass encoding | |
| CN1695381A (en) | Sharpness enhancement in post-processing of digital video signals using coding information and local spatial features | |
| KR20010110629A (en) | Method and system for compressing motion image information | |
| KR20130053645A (en) | Method and apparatus for video encoding/decoding using adaptive loop filter | |
| CN114640847B (en) | Encoding and decoding method, device and equipment thereof | |
| He et al. | Adaptive HEVC video steganography with high performance based on attention-net and PU partition modes | |
| CN113329229A (en) | High-capacity hiding method for H.265 video information with high-efficiency fidelity | |
| CN114125445A (en) | Decoding method, device, equipment and machine readable storage medium | |
| Boitard et al. | Motion-guided quantization for video tone mapping | |
| CN113315976A (en) | Three-in-one high information content embedding method for low-resolution video | |
| JP2008544621A (en) | Encoding and decoding method and apparatus for improving video error concealment | |
| KR100584422B1 (en) | Method and device for compressing image data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |